UNICEF Assignment

Reasons for Leaving School Early - Quarto Report

Report Summary

This report seeks to understand why some children leave school early.

• Data for this analysis and report were sourced from UNICEF
• Please find the Tableau Version of this report here
• Due to some of the graphs not rendering on GitHub Pages, links to these files have been included in their respective explanations. I tried to include the urls within the html file itself, but was unable to position them successfully.

Education

Every child has right to education. It should not even be in question. Unfortunately, this does not translate to them receiving an education globally. Over 600 million children worldwide are unable to attain foundational levels in literacy and mathematics. Children are deprived of an education for many reasons. Poverty remains the key barrier. Children living in economic fragility, political instability, conflicts or natural disasters, are more likely to suffer this fate. Without a quality education, children face considerable barriers to employment in their life. This vicious circle never ends as they then become parents and their children are trapped in the same cycle too. Thus, their ability to shape a better future for themselves and their society seems impossible.

By using unicef_indicator_1 and unicef_metadata, this report aims to look at some factors as to why adolescents of a lower secondary school age, drop out of school. It will also show geographically, which areas are the most affected.

from google.colab import drive
drive.mount('/content/drive')
import polars as pl

Mounted at /content/drive

pl.read_csv('drive/MyDrive/Colab Notebooks/unicef_metadata1.csv',
    infer_schema_length=10000,  # Increase infer_schema_length
    schema_overrides={'Population, total': pl.Float64}, # Or, specify the correct dtype
)
meta = pl.read_csv('drive/MyDrive/Colab Notebooks/unicef_metadata1.csv',
    infer_schema_length=10000,
    schema_overrides={'Population, total': pl.Float64},
)

pl.read_csv('drive/MyDrive/Colab Notebooks/Out-of-school rate for adolescents of lower secondary school age (administrative data).csv')

OOSR = pl.read_csv('drive/MyDrive/Colab Notebooks/Out-of-school rate for adolescents of lower secondary school age (administrative data).csv')

Which Countries Have the Most Adolescents Out of School?

Below is an overview of the world of education and the average obs value (% out-of-school rate). Using a diverging reds map, the countries around the Equator in Africa have the highest obs value, with countries in Central America and Asia having a high obs value too.

Link to Map

import polars as pl
import geopandas as gpd
import plotly.express as px

# Load the Excel file
df = OOSR

# Filter where sex == 'Total'
df_total = df.filter(pl.col("sex") == "Total")

# Group by country and compute average dropout rate
avg_dropout = df_total.group_by("country").agg(pl.col("obs_value").mean().alias("avg_dropout_rate"))

# Convert to pandas for Plotly
avg_df = avg_dropout.to_pandas()

# Create a choropleth map
fig = px.choropleth(
    avg_df,
    locations="country",
    locationmode="country names",
    color="avg_dropout_rate",
    color_continuous_scale="Reds",
    title="Average Dropout Rate per Country (Adolescents)"
)

fig.show()

Is anyone left behind?

When looking at the female:male ratios, it is quite obvious that there is some disparity in each country. No continent, not even Europe, has full equality between the two genders. While in the Tableau report, each individual country is displayed, for further clarity, just the continents were chosen for this rendition of the graph. Africa overall has the highest drop-out rates. However, Oceania has the largest disparity between the two genders, with a much larger percentage of males leaving school earlier than their female counterparts. Overall, globally, it can be assumed that if you are born male, there is a higher chance that you will leave school earlier than if you were female.

!pip install --upgrade polars
!pip install country_converter
!pip install plotnine geopandas
from plotnine import *
import country_converter as coco
import polars as pl

gender_data = (OOSR
               .filter(pl.col("sex") != "Total")
               .group_by(["country","sex"])
               .agg(pl.col("obs_value").median())
               .pivot(values="obs_value",index="country",columns="sex"))

cc = coco.CountryConverter()
countries = gender_data.get_column("country").to_list()
continent_list = cc.convert(names=countries, to='continent', not_found=None)
gender_data_with_continent = gender_data.with_columns(
    pl.Series(continent_list).alias("continent")
)


# Melt the DataFrame for plotting
melted_data = gender_data_with_continent.melt(id_vars=["country", "continent"], variable_name="sex", value_name="obs_value")

# Create the plot using plotnine
continent_medians = (
    melted_data
    .group_by(["continent", "sex"])
    .agg(pl.col("obs_value").median().alias("median_obs_value"))
)

(ggplot(continent_medians, aes(x="continent", y="median_obs_value",fill="sex"))
 + geom_col(position="dodge")
 + labs(x="Continent", y="Median Out of School Rate", fill="Sex")
 + theme(axis_text_x=element_text(rotation=45))
 + theme_minimal()
 + scale_fill_manual(values=["purple","pink"])
)

Requirement already satisfied: polars in /usr/local/lib/python3.11/dist-packages (1.27.1)
Requirement already satisfied: country_converter in /usr/local/lib/python3.11/dist-packages (1.3)
Requirement already satisfied: pandas>=1.0 in /usr/local/lib/python3.11/dist-packages (from country_converter) (2.2.2)
Requirement already satisfied: numpy>=1.23.2 in /usr/local/lib/python3.11/dist-packages (from pandas>=1.0->country_converter) (2.0.2)
Requirement already satisfied: python-dateutil>=2.8.2 in /usr/local/lib/python3.11/dist-packages (from pandas>=1.0->country_converter) (2.8.2)
Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.11/dist-packages (from pandas>=1.0->country_converter) (2025.2)
Requirement already satisfied: tzdata>=2022.7 in /usr/local/lib/python3.11/dist-packages (from pandas>=1.0->country_converter) (2025.2)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.11/dist-packages (from python-dateutil>=2.8.2->pandas>=1.0->country_converter) (1.17.0)
Requirement already satisfied: plotnine in /usr/local/lib/python3.11/dist-packages (0.14.5)
Requirement already satisfied: geopandas in /usr/local/lib/python3.11/dist-packages (1.0.1)
Requirement already satisfied: matplotlib>=3.8.0 in /usr/local/lib/python3.11/dist-packages (from plotnine) (3.10.0)
Requirement already satisfied: pandas>=2.2.0 in /usr/local/lib/python3.11/dist-packages (from plotnine) (2.2.2)
Requirement already satisfied: mizani~=0.13.0 in /usr/local/lib/python3.11/dist-packages (from plotnine) (0.13.3)
Requirement already satisfied: numpy>=1.23.5 in /usr/local/lib/python3.11/dist-packages (from plotnine) (2.0.2)
Requirement already satisfied: scipy>=1.8.0 in /usr/local/lib/python3.11/dist-packages (from plotnine) (1.14.1)
Requirement already satisfied: statsmodels>=0.14.0 in /usr/local/lib/python3.11/dist-packages (from plotnine) (0.14.4)
Requirement already satisfied: pyogrio>=0.7.2 in /usr/local/lib/python3.11/dist-packages (from geopandas) (0.10.0)
Requirement already satisfied: packaging in /usr/local/lib/python3.11/dist-packages (from geopandas) (24.2)
Requirement already satisfied: pyproj>=3.3.0 in /usr/local/lib/python3.11/dist-packages (from geopandas) (3.7.1)
Requirement already satisfied: shapely>=2.0.0 in /usr/local/lib/python3.11/dist-packages (from geopandas) (2.1.0)
Requirement already satisfied: contourpy>=1.0.1 in /usr/local/lib/python3.11/dist-packages (from matplotlib>=3.8.0->plotnine) (1.3.2)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.11/dist-packages (from matplotlib>=3.8.0->plotnine) (0.12.1)
Requirement already satisfied: fonttools>=4.22.0 in /usr/local/lib/python3.11/dist-packages (from matplotlib>=3.8.0->plotnine) (4.57.0)
Requirement already satisfied: kiwisolver>=1.3.1 in /usr/local/lib/python3.11/dist-packages (from matplotlib>=3.8.0->plotnine) (1.4.8)
Requirement already satisfied: pillow>=8 in /usr/local/lib/python3.11/dist-packages (from matplotlib>=3.8.0->plotnine) (11.1.0)
Requirement already satisfied: pyparsing>=2.3.1 in /usr/local/lib/python3.11/dist-packages (from matplotlib>=3.8.0->plotnine) (3.2.3)
Requirement already satisfied: python-dateutil>=2.7 in /usr/local/lib/python3.11/dist-packages (from matplotlib>=3.8.0->plotnine) (2.8.2)
Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.11/dist-packages (from pandas>=2.2.0->plotnine) (2025.2)
Requirement already satisfied: tzdata>=2022.7 in /usr/local/lib/python3.11/dist-packages (from pandas>=2.2.0->plotnine) (2025.2)
Requirement already satisfied: certifi in /usr/local/lib/python3.11/dist-packages (from pyogrio>=0.7.2->geopandas) (2025.1.31)
Requirement already satisfied: patsy>=0.5.6 in /usr/local/lib/python3.11/dist-packages (from statsmodels>=0.14.0->plotnine) (1.0.1)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.11/dist-packages (from python-dateutil>=2.7->matplotlib>=3.8.0->plotnine) (1.17.0)

DeprecationWarning: The argument `columns` for `DataFrame.pivot` is deprecated. It has been renamed to `on`.
<ipython-input-10-033859602bac>:23: DeprecationWarning: `DataFrame.melt` is deprecated. Use `unpivot` instead, with `index` instead of `id_vars` and `on` instead of `value_vars`

Does Wealth Reduce School Dropouts?

We can see the by comparing the average GDP per capita against the average obs value, countries with a High/Middle Income tend to have a lower average obs value. Only the countries that have low income (<$20,000 avg. GDP per capita) have higher average obs. values.

gdp_dropout = (meta.join(OOSR, on="country")
.group_by("country")
.agg([
    pl.col("GDP per capita (constant 2015 US$)").mean().alias("GDP_per_capita"),
    pl.col("obs_value").mean().alias("avg_obs_value")
]))
(ggplot(gdp_dropout, aes(x="avg_obs_value", y="GDP_per_capita"))
+ geom_point()
+ labs(x="Out of School Rate", y="GDP per capita (constant 2015 US$)")
+ theme_seaborn()
)

/usr/local/lib/python3.11/dist-packages/plotnine/layer.py:364: PlotnineWarning: geom_point : Removed 2 rows containing missing values.

Life Expectancy?

This heat map displays the countries aligned from worst to best on the y-axis in terms of their life expectancy. As seen on the right, the colour determines the obs value. There is the familiar cluster of African countries towards the bottom of the graph.

In 2023, the average life expectancy in Sierra Leone was 61.786 years compared to the 83.63 years in Liechtenstein.

While this comparison is bleak, life expectancy in Africa is rising as reported by the World Health Organization. Ringfencing health systems in poorer countries is crucial to the survival and ability for children to stay in school longer. Read more about it in this article linked here.

import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

# Load data
metadata = meta
dropout = OOSR

metadata_pd = metadata.to_pandas()
dropout_pd = dropout.to_pandas()

# Filter dropout data for 'Total' sex
dropout_total = dropout_pd[dropout_pd["sex"] == "Total"]

# Calculate max dropout rate per country
dropout_max = dropout_total.groupby("country")["obs_value"].max().reset_index()
dropout_max.columns = ["country", "max_dropout"]

# Calculate median life expectancy per country
life_expectancy = metadata_pd.groupby("country")["Life expectancy at birth, total (years)"].median().reset_index()
life_expectancy.columns = ["country", "median_life_expectancy"]

# Merge both datasets
merged = pd.merge(dropout_max, life_expectancy, on="country", how="inner")

# Set up heatmap-style DataFrame
heatmap_df = merged.pivot_table(
    index="country",
    values="max_dropout",
    columns="median_life_expectancy"
)

# Sort countries for readability
merged_sorted = merged.sort_values("median_life_expectancy", ascending=False)

# Plot heatmap (using just country as axis for simplicity)
plt.figure(figsize=(10, 8))
sns.heatmap(
    data=merged_sorted.set_index("country")[["max_dropout"]],
    cmap="YlOrRd",
    annot=False,
    fmt=".1f",
    linewidths=0.5,
    cbar_kws={'label': 'Max Dropout Rate (%)'},
    xticklabels=False
)

plt.title("Heatmap of Max Dropout Rate vs Median Life Expectancy by Country")
plt.xlabel("Max Dropout Rate")
plt.ylabel("Country (sorted by life expectancy)")
plt.tight_layout(pad=0.5)
plt.show()

Does Military Expenditure have an effect?

YES!

It is quite obvious from this sunburst graph that low military expenditure does have an effect on a country’s chances of having a higher drop-out rate. Only the countries in the largest sector, Low Income, have the highest drop-out values. None of the countries in the High/Medium Income categories have high obs values.

Education can be a lifeline for children stuck in crises. For children in an emergency such as living in a war-torn country, education is about more than just the right to learn.

Schools provide a safe haven for children to escape the physical danger that constantly surrounds them. In an emergency, education is normally the first service suspended… and the last to be restored.

UNICEF works to provide an uninterrupted education for every child affected by a humanitarian crisis. Learn more about their work here.

Please click this link to view the Military graph

import pandas as pd
import plotly.express as px



# Load data
metadata_pd = meta
dropout_pd = OOSR

metadata_pd = metadata.to_pandas()
dropout_pd = dropout.to_pandas()

# Filter dropout data for 'Total' sex
dropout_total = dropout_pd[dropout_pd["sex"] == "Total"]

# Median dropout rate per country
dropout_median = dropout_total.groupby("country")["obs_value"].median().reset_index()
dropout_median.columns = ["country", "median_dropout"]

# Aggregate metadata
metadata_grouped = metadata_pd.groupby("country").agg({
    "Military expenditure (% of GDP)": "max",
    "Life expectancy at birth, total (years)": "median",
    "GDP per capita (constant 2015 US$)": "max"
}).reset_index()

# Merge
merged = pd.merge(metadata_grouped, dropout_median, on="country", how="inner")
merged = merged.rename(columns={
    "Military expenditure (% of GDP)": "military_expenditure",
    "Life expectancy at birth, total (years)": "life_expectancy",
    "GDP per capita (constant 2015 US$)": "gdp_per_capita"
})

# Define income level
def income_level(gdp):
    if gdp >= 45000:
        return "High Income"
    elif gdp >= 20000:
        return "Middle Income"
    else:
        return "Low Income"

merged["income_level"] = merged["gdp_per_capita"].apply(income_level)

# Clean up
merged = merged.dropna(subset=["military_expenditure", "median_dropout", "life_expectancy"])

# Sunburst chart
fig = px.sunburst(
    merged,
    path=["income_level", "country"],
    values="military_expenditure",
    color="median_dropout",
    color_continuous_scale="YlOrRd",
    hover_data=["life_expectancy"],
    labels={
        "military_expenditure": "Military Expenditure (% of GDP)",
        "median_dropout": "Dropout Rate (%)",
        "life_expectancy": "Life Expectancy (Years)",
        "income_level": "Income Level"
    },
    title="Military Expenditure Breakdown by Income Group<br>(Dropout Rate = Color, Life Expectancy on Hover)"
)



fig.update_traces(insidetextorientation='radial')
fig.show()

Life Expectancy vs Birth Rate… and more

This graph is one of the most telling of this entire report. On the y-axis is the maximum life expectancy and on the x-axis is the maximum birth rate (per 1,000 people).

Each circle represents a country’s income level and the size dictates its median obs value. The larger the circle, the higher the median value.

The three colours are as described in the legend.

So what is it telling us?

• As a country’s birth rate increases, its life expectancy decreases
• When this happens, the median obs value increases
• Judging by the size of the circles, most low income countries have a higher median obs value than medium/high income countries.

Income (USD) is divided by: High > 45,000, Medium >= 20,000, Low < 20,000

Please click this link to view the Life Expectancy vs Birth Rate graph

import plotly.express as px
import pandas as pd

# Load metadata and dropout files
metadata = meta
dropout = OOSR

# Convert Polars DataFrames to pandas DataFrames
metadata_pd = metadata.to_pandas()
dropout_pd = dropout.to_pandas()

# Filter dropout data for 'Total' sex
dropout_total = dropout_pd[dropout_pd["sex"] == "Total"]  # Use pandas filtering

# Calculate median dropout rate per country
dropout_median = dropout_total.groupby("country")["obs_value"].median().reset_index()
dropout_median.columns = ["country", "median_dropout"]

# Get max birth rate and life expectancy per country using pandas
metadata_grouped = metadata_pd.groupby("country").agg({
    "Birth rate, crude (per 1,000 people)": "max",
    "Life expectancy at birth, total (years)": "max",
    "GDP per capita (constant 2015 US$)": "max"
}).reset_index()

# Merge with dropout data (both are now pandas DataFrames)
merged = pd.merge(metadata_grouped, dropout_median, on="country", how="inner")

# Rename for easier access
merged = merged.rename(columns={
    "Birth rate, crude (per 1,000 people)": "birth_rate",
    "Life expectancy at birth, total (years)": "life_expectancy",
    "GDP per capita (constant 2015 US$)": "gdp_per_capita"
})


# Create income level calculated field
def income_level(gdp):
    if gdp >= 45000:
        return "High Income"
    elif gdp >= 20000:
        return "Middle Income"
    else:
        return "Low Income"

merged["income_level"] = merged["gdp_per_capita"].apply(income_level)

# Color mapping
color_map = {
    "High Income": "green",
    "Middle Income": "orange",
    "Low Income": "gold"
}
merged["color"] = merged["income_level"].map(color_map)

fig = px.scatter(
    merged,
    x="birth_rate",
    y="life_expectancy",
    size="median_dropout",
    color="income_level",
    color_discrete_map={  # Custom color mapping
        "High Income": "green",
        "Middle Income": "orange",
        "Low Income": "blue"
    },
    hover_name="country",  # Show country name on hover
    hover_data={  # Show additional data on hover
        "birth_rate": True,
        "life_expectancy": True,
        "median_dropout": True,
        "income_level": True
    },
    title="Max Birth Rate vs. Life Expectancy by Country (Interactive)<br>(Trendline: Linear Regression)",
    labels={
        "birth_rate": "Max Birth Rate (per 1,000 people)",
        "life_expectancy": "Max Life Expectancy (years)",
        "median_dropout": "Median Dropout Rate",
        "income_level": "Income Level"
    },
    trendline="ols"
)

fig.show()

Concluding Thoughts

In conclusion, the analysis of why adolescents of lower-secondary school age drop-out of school early has revealed some consistent challenges. While levels of education are increasing globally, disparities still remain, particularly in Africa and Oceania.

To address these challenges, concerted efforts will be needed for future education investment and agreements. Collaboration between governments of both poor and wealthy countries, international organisations, and local communities will be key for achieving universal access to a longer period in education.

Only by working together, can we as one people make significant strides towards a future where education is a fundamental human right for all.